FPGA Platform applied for Facial Expression Recognition System using Convolutional Neural Networks

Phan-Xuan, Hanh; Le-Tien, Thuong; Nguyen-Tan, Sy

doi:10.1016/j.procs.2019.04.087

Cited by 21 publications

(13 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As shown in Table 4, The proposed CNN accelerator was compared to previous implementations. Compared with the method of [40], the new Type-2 method was found to be more accurate, to use fewer resources, and to have frame rate approximately 25 times faster. Compared to the work in [41], the new Type-2 approach is more accurate, uses 2.2 times fewer DSPs, and has a frame rate about 20 times faster.…”

Section: Comparison With Other Lightweight Cnnsmentioning

confidence: 94%

Section: Fpga-based Cnn Accelerator For Facial Emotion Recognitionmentioning

confidence: 99%

“…Due to these advantages, several FPGA-based CNN accelerators for facial emotion recognition have been proposed [40]- [44]. Phan-Xuan et al [40] implemented a CNN accelerator on a Xilinx Zynq-XC7Z020 FPGA using high-level-synthesis (HLS). They designed the CNN accelerator that stores the results and inputs of the layer in DRAM to reduce block RAM (BRAM) usage, and used VDMA for fast DRAM access.…”

Section: Fpga-based Cnn Accelerator For Facial Emotion Recognitionmentioning

confidence: 99%

“…In previous works [40]- [42], a simple CNN architecture was implemented involving fully-connected, convolution, and pooling layers. Its number of parameters and computational complexity was low, but so was its accuracy, making it unsuitable for applications requiring highly accurate facial emotion recognition.…”

Section: Fpga-based Cnn Accelerator For Facial Emotion Recognitionmentioning

confidence: 99%

See 3 more Smart Citations

A Resource Efficient Integer-Arithmetic-Only FPGA-Based CNN Accelerator for Real-Time Facial Emotion Recognition

2021

View full text Add to dashboard Cite

Recently, much research has been conducted on the recognition of facial emotion using convolutional neural networks (CNNs), which show excellent performance in computer vision. To obtain a high classification accuracy, a CNN architecture with many parameters and high computational complexity is required. However, this is not suitable for embedded systems where hardware resources are limited. In this paper, we present a lightweight CNN architecture optimized for embedded systems. The proposed CNN architecture has a small memory footprint and low computational complexity. Furthermore, a novel hardware-friendly quantization method that uses only integer-arithmetic is proposed. The proposed hardwarefriendly quantization method maps the scale factors to power-of-two terms and replaces multiplication and division operations using scale factors with shift operations. To improve the generalization and classification performance of the CNN, we create the FERPlus-A dataset. This is a new training dataset created using a variety of image processing algorithms. After training with FERPlus-A, quantization has been performed. The size of a quantized CNN parameter is about 0.39 MB, and the number of operations is about 28 M integer operations (IOPs). By evaluating the performance of the quantized CNN that uses only integer-arithmetic on the FERPlus test dataset, the classification accuracy is approximately 86.58%. It achieved higher accuracy than other lightweight CNNs in prior studies. The proposed CNN architecture that uses only integerarithmetic is implemented on the Xilinx ZC706 SoC platform for real-time facial emotion recognition by applying parallelism strategies and efficient data caching strategies. The FPGA-based CNN accelerator implemented for real-time facial emotion recognition achieves about 10 frames per second (FPS) at 250 MHz and consumes 2.3 W.

show abstract

Section: Comparison With Other Lightweight Cnnsmentioning

confidence: 94%

Section: Fpga-based Cnn Accelerator For Facial Emotion Recognitionmentioning

confidence: 99%

Section: Fpga-based Cnn Accelerator For Facial Emotion Recognitionmentioning

confidence: 99%

Section: Fpga-based Cnn Accelerator For Facial Emotion Recognitionmentioning

confidence: 99%

See 2 more Smart Citations

A Resource Efficient Integer-Arithmetic-Only FPGA-Based CNN Accelerator for Real-Time Facial Emotion Recognition

2021

View full text Add to dashboard Cite

show abstract

“…The welldesigned weights and thresholds of the neural network can be therefore arrived at. And we can embed this algorithm into integrated devices to measure velocity excellently in real time [24], [25]. Finally, the effectiveness of the proposed method will be experimentally verified by measuring the detecting errors of the velocity under the temperature rise condition.…”

Section: Introductionmentioning

confidence: 95%

Extending Velocity Sensor Bandwidth by Compensating Temperature Dependency Based on BP Neural Network

et al. 2019

View full text Add to dashboard Cite

A compensation method for a magnetoelectric velocity sensor (MVS) is always necessary, which can lower the resonance frequency of the measuring system and subsequently extend the measuring bandwidth. In this paper, a novel compensation method is proposed based on the BP neural network under the TensorFlow architecture. Comparing with the existing methods, the new method does not depend upon an accurate model of the MVS any more, whose parameters are badly influenced by the temperature. The dynamic compensator is connected with the sensor. The BP neural network algorithm is used to identify compensation parameters. The dynamic compensator works at state of the optimum parameter all the time to compensate the dynamic performance of MVS by training the weights and thresholds of the neural network. The experiment results show that velocity measurement deviation is within ±5% error band by the dynamic compensator, which can reduce the measurement deviation caused by the variation of temperature and improve the measurement accuracy. The bandwidth can be as low as 0.28Hz. The dynamic compensator is superior to Random Forests and RBF Neutral Network in implement in FPGA/CPLD. Its' accuracy is superior to zero-pole compensation method. This method leads a new way to weaken the temperature variation characteristics of the velocity sensor and improve the measurement performance. INDEX TERMS Velocity sensor, bandwidth expansion, temperature dependence, BP neural network.

show abstract

A DCNN Based Real-Time Authentication System Using Facial Emotions

James¹,

Mun

Ritta³

2021

Intelligent Systems Reference Library

View full text Add to dashboard Cite

FPGA Platform applied for Facial Expression Recognition System using Convolutional Neural Networks

Cited by 21 publications

References 7 publications

A Resource Efficient Integer-Arithmetic-Only FPGA-Based CNN Accelerator for Real-Time Facial Emotion Recognition

A Resource Efficient Integer-Arithmetic-Only FPGA-Based CNN Accelerator for Real-Time Facial Emotion Recognition

Extending Velocity Sensor Bandwidth by Compensating Temperature Dependency Based on BP Neural Network

A DCNN Based Real-Time Authentication System Using Facial Emotions

Contact Info

Product

Resources

About